A Unit Touch Gesture Model of Performance Time Prediction for Mobile Devices

Abstract

Objectiv e:The aim of this study is to propose a unit touch gesture model, which would be useful to predict the performance time on mobile devices. Background:When estimating usability based on Model-based Evaluation (MBE) in interfaces, the GOMS model measured 'operators' to predict the execution time in the desktop environment. Therefore, this study used the concept of operator in GOMS for touch gestures. Since the touch gestures are comprised of possible unit touch gestures, these unit touch gestures can predict to performance time with unit touch gestures on mobile devices. Method:In order to extract unit touch gestures, manual movements of subjects were recorded in the 120 fps with pixel coordinates. Touch gestures are classified with 'out of range', 'registration', 'continuation' and 'termination' of gesture. Results:As a results, six unit touch gestures were extracted, which are hold down (H), Release (R), Slip (S), Curved-stroke (Cs), Path-stroke (Ps) and Out of range (Or). The movement time predicted by the unit touch gesture model is not significantly different from the participants' execution time. The measured six unit touch gestures can predict movement time of undefined touch gestures like user-defined gestures. Conclusion:In conclusion, touch gestures could be subdivided into six unit touch gestures. Six unit touch gestures can explain almost all the current touch gestures including user-defined gestures. So, this model provided in this study has a high predictive power. The model presented in the study could be utilized to predict the performance time of touch gestures. Application:The unit touch gestures could be simply added up to predict the performance time without measuring the performance time of a new gesture.

A novel approach for usability evaluation of mobile applications

“…First, we assumed a linear relationship between the TCT and efficiency. Second, we assumed that users' skill levels and distraction could be categorized into three levels based on prior studies in this domain (Holleis et al, 2007;Kim & Myung, 2016;Nyström, 2018). In our future study, we plan to compare the findings of this approach with human-subject experiments to further adjust the assumptions and parameters.…”

“…Recently, some CPM approaches were used to evaluate the usability of graphical user interfaces (GUI) of mobile apps by measuring task completion time (TCT). For example, Goals, Operators, Methods, and Selection rules (GOMS) (Card et al, 1980) method was used to evaluate the TCT of a chatbot (Prahara & Saputro, 2019), gesture-based touchscreen input interface (Kim & Myung, 2016;Nyström, 2018), adaptive webpage (Rebai et al, 2016), security interface (Din, 2015), and website navigation on a smartphone (Jung & Jang, 2015). Other CPM methods such as Queueing Network -Model Human Processor (QN-MHP) (Liu et al, 2006) was also used to measure TCT and mental workload, in performing tasks such as touchscreen drag gesture (Jeong & Liu, 2019), transcription typing on a touchscreen (Cao et al, 2018), and finger swipe gestures (Jeong & Liu, 2016).…”

“…The GLM was developed based on previous human-inthe-loop studies (Kim & Myung, 2016;Nyström, 2018) in order to calculate the TCT with uncertainty considering user skill level and distraction. The GLM follows the same logic as in the Keystroke-Level Model (Card et al, 1980); but includes additional gesture-specific operators such as drag, flick, tap, double-tap, zoom-in, zoom-out, or pointing, which are especially helpful in evaluation of mobile apps.…”

“…The GLM follows the same logic as in the Keystroke-Level Model (Card et al, 1980); but includes additional gesture-specific operators such as drag, flick, tap, double-tap, zoom-in, zoom-out, or pointing, which are especially helpful in evaluation of mobile apps. For estimating the TCT of each operator, this study used the average TCTs from prior studies (Kim & Myung, 2016;Nyström, 2018) as the "most-likely" TCT, plus and minus of two standard deviations as the "worst" and "best" TCT to better represent uncertainties with different operator's skill levels and account for the true majority of the population based on a normal curve (i.e., the 'best' performer accomplishes a task faster and the 'worst' performer has the longest task completion time). Furthermore, considering unpredictable environment of a mobile phone user, three levels of distractions (i.e., no distraction, slight distraction (e.g., standing in a street), and strong distraction (e.g., walking in a busy street)) were included and the TCT was estimated depending on the distraction level (Holleis et al, 2007;Nyström, 2018).…”

Estimating Cognitive Load in a Mobile Personal Health Record Application: A Cognitive Task Analysis Approach

“…However, since the KLM was developed for desktop computer interactions, it cannot be applied to evaluate the use of smart devices, which have become increasingly prevalent in our lives due to technological advancements. Recognizing this limitation, researchers have proposed new models [19,[21][22][23][24][25] based on the KLM method specifically tailored for mobile devices, aimed at assessing interactions with emerging technology.…”